1 /*****************************************************************************
2 * audio.c: audio decoder using libavcodec library
3 *****************************************************************************
4 * Copyright (C) 1999-2003 VLC authors and VideoLAN
7 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
8 * Gildas Bazin <gbazin@videolan.org>
10 * This program is free software; you can redistribute it and/or modify it
11 * under the terms of the GNU Lesser General Public License as published by
12 * the Free Software Foundation; either version 2.1 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public License
21 * along with this program; if not, write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
23 *****************************************************************************/
25 /*****************************************************************************
27 *****************************************************************************/
32 #include <vlc_common.h>
34 #include <vlc_codec.h>
35 #include <vlc_avcodec.h>
37 #include <libavcodec/avcodec.h>
38 #include <libavutil/mem.h>
40 #include <libavutil/audioconvert.h>
44 /*****************************************************************************
45 * decoder_sys_t : decoder descriptor
46 *****************************************************************************/
49 AVCODEC_COMMON_MEMBERS
51 /* Temporary buffer for libavcodec */
58 audio_sample_format_t aout_format;
72 int pi_extraction[AOUT_CHAN_MAX];
73 int i_previous_channels;
74 int64_t i_previous_layout;
77 #define BLOCK_FLAG_PRIVATE_REALLOCATED (1 << BLOCK_FLAG_PRIVATE_SHIFT)
79 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust );
81 static void InitDecoderConfig( decoder_t *p_dec, AVCodecContext *p_context )
83 if( p_dec->fmt_in.i_extra > 0 )
85 const uint8_t * const p_src = p_dec->fmt_in.p_extra;
89 if( p_dec->fmt_in.i_codec == VLC_CODEC_FLAC )
92 i_size = p_dec->fmt_in.i_extra - 8;
94 else if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
96 static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
97 /* Find alac atom XXX it is a bit ugly */
98 for( i_offset = 0; i_offset < p_dec->fmt_in.i_extra - sizeof(p_pattern); i_offset++ )
100 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
103 i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
110 i_size = p_dec->fmt_in.i_extra;
115 p_context->extradata =
116 malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
117 if( p_context->extradata )
119 uint8_t *p_dst = p_context->extradata;
121 p_context->extradata_size = i_size;
123 memcpy( &p_dst[0], &p_src[i_offset], i_size );
124 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
130 p_context->extradata_size = 0;
131 p_context->extradata = NULL;
135 /*****************************************************************************
136 * InitAudioDec: initialize audio decoder
137 *****************************************************************************
138 * The avcodec codec will be opened, some memory allocated.
139 *****************************************************************************/
140 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
141 AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
143 decoder_sys_t *p_sys;
145 /* Allocate the memory needed to store the decoder's structure */
146 if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
151 p_codec->type = AVMEDIA_TYPE_AUDIO;
152 p_context->codec_type = AVMEDIA_TYPE_AUDIO;
153 p_context->codec_id = i_codec_id;
154 p_sys->p_context = p_context;
155 p_sys->p_codec = p_codec;
156 p_sys->i_codec_id = i_codec_id;
157 p_sys->psz_namecodec = psz_namecodec;
158 p_sys->b_delayed_open = true;
160 // Initialize decoder extradata
161 InitDecoderConfig( p_dec, p_context);
163 /* ***** Open the codec ***** */
164 if( ffmpeg_OpenCodec( p_dec ) < 0 )
166 msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
167 free( p_sys->p_context->extradata );
174 case CODEC_ID_WAVPACK:
175 p_sys->i_output_max = 8 * sizeof(int32_t) * 131072;
178 p_sys->i_output_max = p_sys->p_context->channels * sizeof(int32_t) * p_sys->p_context->sample_rate * 2;
181 p_sys->i_output_max = 8 * sizeof(int32_t) * 65535;
183 case CODEC_ID_WMAPRO:
184 p_sys->i_output_max = 8 * sizeof(float) * 6144; /* (1 << 12) * 3/2 */
187 p_sys->i_output_max = 0;
190 if( p_sys->i_output_max < AVCODEC_MAX_AUDIO_FRAME_SIZE )
191 p_sys->i_output_max = AVCODEC_MAX_AUDIO_FRAME_SIZE;
192 msg_Dbg( p_dec, "Using %d bytes output buffer", p_sys->i_output_max );
193 p_sys->p_output = av_malloc( p_sys->i_output_max );
195 p_sys->p_samples = NULL;
196 p_sys->i_samples = 0;
197 p_sys->i_reject_count = 0;
198 p_sys->b_extract = false;
199 p_sys->i_previous_channels = 0;
200 p_sys->i_previous_layout = 0;
203 p_dec->fmt_out.i_cat = AUDIO_ES;
204 /* Try to set as much information as possible but do not trust it */
205 SetupOutputFormat( p_dec, false );
207 date_Set( &p_sys->end_date, 0 );
208 if( p_dec->fmt_out.audio.i_rate )
209 date_Init( &p_sys->end_date, p_dec->fmt_out.audio.i_rate, 1 );
210 else if( p_dec->fmt_in.audio.i_rate )
211 date_Init( &p_sys->end_date, p_dec->fmt_in.audio.i_rate, 1 );
216 /*****************************************************************************
217 * SplitBuffer: Needed because aout really doesn't like big audio chunk and
218 * wma produces easily > 30000 samples...
219 *****************************************************************************/
220 static block_t *SplitBuffer( decoder_t *p_dec )
222 decoder_sys_t *p_sys = p_dec->p_sys;
223 int i_samples = __MIN( p_sys->i_samples, 4096 );
226 if( i_samples == 0 ) return NULL;
228 if( ( p_buffer = decoder_NewAudioBuffer( p_dec, i_samples ) ) == NULL )
231 p_buffer->i_pts = date_Get( &p_sys->end_date );
232 p_buffer->i_length = date_Increment( &p_sys->end_date, i_samples )
235 if( p_sys->b_extract )
236 aout_ChannelExtract( p_buffer->p_buffer, p_dec->fmt_out.audio.i_channels,
237 p_sys->p_samples, p_sys->p_context->channels, i_samples,
238 p_sys->pi_extraction, p_dec->fmt_out.audio.i_bitspersample );
240 memcpy( p_buffer->p_buffer, p_sys->p_samples, p_buffer->i_buffer );
242 p_sys->p_samples += i_samples * p_sys->p_context->channels * ( p_dec->fmt_out.audio.i_bitspersample / 8 );
243 p_sys->i_samples -= i_samples;
248 /*****************************************************************************
249 * DecodeAudio: Called to decode one frame
250 *****************************************************************************/
251 block_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
253 decoder_sys_t *p_sys = p_dec->p_sys;
254 int i_used, i_output;
259 if( !pp_block || !*pp_block ) return NULL;
263 if( !p_sys->p_context->extradata_size && p_dec->fmt_in.i_extra &&
264 p_sys->b_delayed_open)
266 InitDecoderConfig( p_dec, p_sys->p_context);
267 if( ffmpeg_OpenCodec( p_dec ) )
268 msg_Err( p_dec, "Cannot open decoder %s", p_sys->psz_namecodec );
270 if( p_sys->b_delayed_open )
272 block_Release( p_block );
276 if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
278 block_Release( p_block );
279 avcodec_flush_buffers( p_sys->p_context );
280 p_sys->i_samples = 0;
281 date_Set( &p_sys->end_date, 0 );
283 if( p_sys->i_codec_id == CODEC_ID_MP2 || p_sys->i_codec_id == CODEC_ID_MP3 )
284 p_sys->i_reject_count = 3;
288 if( p_sys->i_samples > 0 )
291 p_buffer = SplitBuffer( p_dec );
292 if( !p_buffer ) block_Release( p_block );
296 if( !date_Get( &p_sys->end_date ) && !p_block->i_pts )
298 /* We've just started the stream, wait for the first PTS. */
299 block_Release( p_block );
303 if( p_block->i_buffer <= 0 )
305 block_Release( p_block );
309 if( (p_block->i_flags & BLOCK_FLAG_PRIVATE_REALLOCATED) == 0 )
311 *pp_block = p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
314 p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
315 memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
317 p_block->i_flags |= BLOCK_FLAG_PRIVATE_REALLOCATED;
322 i_output = __MAX( p_block->i_buffer, p_sys->i_output_max );
323 if( i_output > p_sys->i_output_max )
325 /* Grow output buffer if necessary (eg. for PCM data) */
326 p_sys->p_output = av_realloc( p_sys->p_output, i_output );
329 av_init_packet( &pkt );
330 pkt.data = p_block->p_buffer;
331 pkt.size = p_block->i_buffer;
332 i_used = avcodec_decode_audio3( p_sys->p_context,
333 (int16_t*)p_sys->p_output, &i_output,
336 if( i_used < 0 || i_output < 0 )
339 msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
342 block_Release( p_block );
345 else if( (size_t)i_used > p_block->i_buffer )
347 i_used = p_block->i_buffer;
350 p_block->i_buffer -= i_used;
351 p_block->p_buffer += i_used;
353 } while( p_block->i_buffer > 0 && i_output <= 0 );
355 if( p_sys->p_context->channels <= 0 || p_sys->p_context->channels > 8 ||
356 p_sys->p_context->sample_rate <= 0 )
358 msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
359 p_sys->p_context->channels, p_sys->p_context->sample_rate );
360 block_Release( p_block );
364 if( p_dec->fmt_out.audio.i_rate != (unsigned int)p_sys->p_context->sample_rate )
366 date_Init( &p_sys->end_date, p_sys->p_context->sample_rate, 1 );
367 date_Set( &p_sys->end_date, p_block->i_pts );
370 /* **** Set audio output parameters **** */
371 SetupOutputFormat( p_dec, true );
373 if( p_block->i_pts != 0 &&
374 p_block->i_pts != date_Get( &p_sys->end_date ) )
376 date_Set( &p_sys->end_date, p_block->i_pts );
380 /* **** Now we can output these samples **** */
381 p_sys->i_samples = i_output / (p_dec->fmt_out.audio.i_bitspersample / 8) / p_sys->p_context->channels;
382 p_sys->p_samples = p_sys->p_output;
384 /* Silent unwanted samples */
385 if( p_sys->i_reject_count > 0 )
387 memset( p_sys->p_output, 0, i_output );
388 p_sys->i_reject_count--;
391 p_buffer = SplitBuffer( p_dec );
392 if( !p_buffer ) block_Release( p_block );
396 /*****************************************************************************
397 * EndAudioDec: audio decoder destruction
398 *****************************************************************************/
399 void EndAudioDec( decoder_t *p_dec )
401 decoder_sys_t *p_sys = p_dec->p_sys;
403 av_free( p_sys->p_output );
406 /*****************************************************************************
408 *****************************************************************************/
410 void GetVlcAudioFormat( vlc_fourcc_t *pi_codec, unsigned *pi_bits, int i_sample_fmt )
412 switch( i_sample_fmt )
414 case AV_SAMPLE_FMT_U8:
415 *pi_codec = VLC_CODEC_U8;
418 case AV_SAMPLE_FMT_S32:
419 *pi_codec = VLC_CODEC_S32N;
422 case AV_SAMPLE_FMT_FLT:
423 *pi_codec = VLC_CODEC_FL32;
426 case AV_SAMPLE_FMT_DBL:
427 *pi_codec = VLC_CODEC_FL64;
431 case AV_SAMPLE_FMT_S16:
433 *pi_codec = VLC_CODEC_S16N;
439 static const uint64_t pi_channels_map[][2] =
441 { AV_CH_FRONT_LEFT, AOUT_CHAN_LEFT },
442 { AV_CH_FRONT_RIGHT, AOUT_CHAN_RIGHT },
443 { AV_CH_FRONT_CENTER, AOUT_CHAN_CENTER },
444 { AV_CH_LOW_FREQUENCY, AOUT_CHAN_LFE },
445 { AV_CH_BACK_LEFT, AOUT_CHAN_REARLEFT },
446 { AV_CH_BACK_RIGHT, AOUT_CHAN_REARRIGHT },
447 { AV_CH_FRONT_LEFT_OF_CENTER, 0 },
448 { AV_CH_FRONT_RIGHT_OF_CENTER, 0 },
449 { AV_CH_BACK_CENTER, AOUT_CHAN_REARCENTER },
450 { AV_CH_SIDE_LEFT, AOUT_CHAN_MIDDLELEFT },
451 { AV_CH_SIDE_RIGHT, AOUT_CHAN_MIDDLERIGHT },
452 { AV_CH_TOP_CENTER, 0 },
453 { AV_CH_TOP_FRONT_LEFT, 0 },
454 { AV_CH_TOP_FRONT_CENTER, 0 },
455 { AV_CH_TOP_FRONT_RIGHT, 0 },
456 { AV_CH_TOP_BACK_LEFT, 0 },
457 { AV_CH_TOP_BACK_CENTER, 0 },
458 { AV_CH_TOP_BACK_RIGHT, 0 },
459 { AV_CH_STEREO_LEFT, 0 },
460 { AV_CH_STEREO_RIGHT, 0 },
463 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
465 decoder_sys_t *p_sys = p_dec->p_sys;
467 GetVlcAudioFormat( &p_dec->fmt_out.i_codec,
468 &p_dec->fmt_out.audio.i_bitspersample,
469 p_sys->p_context->sample_fmt );
470 p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
473 if( p_sys->i_previous_channels == p_sys->p_context->channels &&
474 p_sys->i_previous_layout == p_sys->p_context->channel_layout )
478 p_sys->i_previous_channels = p_sys->p_context->channels;
479 p_sys->i_previous_layout = p_sys->p_context->channel_layout;
483 * FIXME should we use fmt_in.audio.i_physical_channels or not ?
485 const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
486 uint32_t pi_order_src[i_order_max];
487 int i_channels_src = 0;
489 if( p_sys->p_context->channel_layout )
491 for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
493 if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
494 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
499 /* Create default order */
501 msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
502 for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
504 if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
505 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
508 if( i_channels_src != p_sys->p_context->channels && b_trust )
509 msg_Err( p_dec, "Channel layout not understood" );
511 uint32_t i_layout_dst;
513 p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
514 &i_layout_dst, &i_channels_dst,
515 NULL, pi_order_src, i_channels_src );
516 if( i_channels_dst != i_channels_src && b_trust )
517 msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
519 p_dec->fmt_out.audio.i_physical_channels =
520 p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
521 p_dec->fmt_out.audio.i_channels = i_channels_dst;